Unless you’re very methodical on test system configurations you’ll find this to be an issue, building an image library was something I also found helpful compared to chasing after every new box that arrived in the test lab (especially notebooks) to make sure each and every UI setting was where it needed to be.
After building a control image library, I simply threw each possible image for a control at it, adds a few seconds to run time, but that’s better then an outright failure fault eh?
While building these tools/libraries takes more time, it’s time very well invested. My scripts would run with a much higher success rate on new test boxes after I adopted that methodology.
that’s my take on it anyway, your results will vary according to your own goals and programming styles